Let's get back on topic.. [edit- this pretty much just expands on what most just posted above]
Always good to define the terms. To clarify, "on-axis", may refer to either the speaker creating the sound or a microphones used to record it.
"Directly on-axis" generally equates to a line down the center of some effective coverage window within which response and sensitivity is held more or less constant. It's is easy to visually identify since its typically directly in-line with the font of the device. The width of the effective coverage window depends on polar pattern and the quality of the speaker or microphone.
The main thing we are concerned about is how much overlap there is between the coverage widow of one or both PA speakers, how much overlap there is between the coverage windows of each of the two microphones, and how much overlap there is between the combined coverage of one or both PA speakers and the combined coverage of the microphone pair.
We don't need to be directly on-axis with the PA speakers themselves, only within their effective coverage window. We don't need to point each microphone directly on-axis with the PA speakers, we only need the PA speaker to be located within the good coverage window of the microphone. We don't have to setup within the effective coverage window of both PA speakers, but the stereo aspects will usually be improved when we can arrange that, and the the stereo aspects will be better when we arrange for an optimal amount of overlap between the effective coverage patterns of the two microphones.
Less critical but also of importance is the offset within and between all those various angles of coverage. We might a find different individual PA speakers to have a wider or narrower angle of effective coverage, and differing degrees of overlapping PA coverage depending on the PA setup and the recording position in the room. Similarly at the pickup end we have wider or narrower windows of effective off-axis coverage from the individual microphones and use them in various stereo configurations that provide different amounts of stereo pattern overlap.
As tapers, we don't have a hand in the design and setup of the PA, but the folks who do very much think in terms of "reverse PAS", at least in effect if not in terminology. The configuration of the PA is arranged to direct its effective coverage window over as much of the audience as practical, essentially pointing the PA directly at you in the audience, while directing as little sound as practical in all other directions.
"PAS" refers specifically to orientation of our microphones from a position out in the audience - angling the stereo pair so that the primary axis of one microphone is pointed directly at the PA speaker on one side of the stage and the the primary axis of the other microphone is pointed directly at the PA on the other side of the stage. We can use alternate stereo microphone configurations that still keep the PA within the effective stereo coverage angle of the stereo pair - that's the critical thing - which may sound better or worse. But none of them will maximize pickup of direct sound from the PA while simultaneously minimizing pickup of sound arriving from everywhere else as much as possible as pointing the microphones directly at the PA speakers is going to do (at least potentially, depending on polar pattern and mics used). When that's the most critical thing to achieve to get a good recording, PAS is the only road that leads all the way to Rome.
